Exposure adjustment accuracy and color spaces.

[Guillaume5FC6]

Hi, I noticed SRGB, Adobe 1998 and Prophoto don't give the same results (especially in the darkest dark) when I modify my images with an exposure adjustment.

Is one of the color space where the results are more physically accurate?

Thanks

[Derek Cross]

Others may disagree, but I suggest unless you have some knowledge of colour management, I suggest in the meantime keep to sRGB.

[Guillaume5FC6]

Thanks Derek, but I do have color management knowledge. I really would want to know the math behind thoses differences. Thanks!

[TheDigitalDog]

Something on your end is off, the three color spaces shouldn't have any such effect reported. We may need to see a screen capture of your color settings, you may need to test this on a color reference image too but what you report should not be happening. 

[Guillaume5FC6]

The test is pretty simple.

Create 3 documents in their appropriate color spaces with a flat color of 50L in the color cielab color picker

-Adobe 1998

-Srgb

-Prophoto

So far the 3 grays are perceptually identical and with the L*A*B* color picker

In each of those documents add a -2 exposure adjustment and a second at -4.

Convert them all to srgb for comparison in one document.

And voilà

[Derek Cross]

Is your monitor capable of rendering these color spaces?

[Guillaume5FC6]

Unfortunately, it's not about the monitor, it's about how the tone mapping of each color space and the different gamma equations they use. Tone mapping is different between color spaces but the math of exposure should be the same.

[NB, colourmanagement net]

"Exposure" adjustment may not be so simple as brightness or levels / curves. Have you tried those ways of adjusting the same files?

Could it be that there's some A.I. in "Exposure" and it's seeing the spread of values differently between those colour spaces? One would expect PS to work on the underlying L*a*b* values when calculating corrections, but, maybe, in the case of "Exposure" adjustment, it doesn’t. 

https://helpx.adobe.com/photoshop/using/color-adjustments.html#adjustments_panel_overview

"Exposure command

Adjusts tonality by performing calculations in a linear color space. Exposure is primarily for use in HDR images. See Adjust Exposure for HDR images."

Given what's written there perhaps it's not behaving exactly as one might expect? Bug?

I hope this helps
neil barstow, colourmanagement net :: adobe forum volunteer:: Co-Author: Getting Colour Right
google me "neil barstow colourmanagement" for lots of free articles on colour management

[Guillaume5FC6]

Hi Neil,

Thanks for your input, If the data are converted to linear before doing the exposure adjustment, eveything should be coherent because most of the differences happen when calculations are done in 0-255 format.

If you do the same exercise in 32bit mode with a 50 Lightness L*A*B* flat color, the results once converted to 16 or 8bit (using exposure 0 and gamma 1 ) yield the same numbers as the Adobe1998 color space.

[TheDigitalDog]

The data is not “converted to linear” in PS. 

[Guillaume5FC6]

Ok interesting, care to explain the difference between 32bit/ channel mode and a linear light mode?

[TheDigitalDog]

One has nothing to do with the other. Or the difference you report.

[Guillaume5FC6]

According to them:

"High dynamic range (HDR) images open up a world of possibilities because they can represent the entire dynamic range of the visible world. Because all the luminance values in a real-world scene are represented proportionately and stored in an HDR image, adjusting the exposure of an HDR image is like adjusting the exposure when photographing a scene in the real world."

https://helpx.adobe.com/photoshop/using/high-dynamic-range-images.html

[TheDigitalDog]

Now you're off to HDR?

[TheDigitalDog]

They have different gamma encoding, so what? 

[Guillaume5FC6]

It points to be something more than that. 😞

[D Fosse]

This is normal and expected. There is nothing wrong, and they're all accurate.

All numerical adjustments are color space specific.

The numbers mean different things in different color spaces. So any given numerical adjustment will have different effects. If you drag an adjustment layer from an sRGB file over on a ProPhoto file, you will see that the result can be dramatically different.

[Guillaume5FC6]

@D Fosse  Thank you,

2nd part of my question is about perception. I'll try to be clear as english is not my first language. So do not hesitate to ask me to clarify if it seems confusing. 

-If we had a hypothetical perfect true 8bit monitor.

-That we created a gray square of  x lightness in 3 documents in their respective color space. (Srgb, Adobe 1998, ProPhoto)

-In each those documents we masked half the square and applied an exposure adjustment of -2 stops

-Then if we went on with a spot meter to mesure the light emitted, would we get a -2 stop difference within our squares?

-Would the contrast be the same accross those 3 color spaces once it is displaid? 

I understand it might seem specific and unusual. But I'm a painter and we can mesure contrast in reflected light between  physical gray pigments with a light meter. In the end, I would like to use the most perceptually "correct" color space to work in 8bit with exposure as a measurement tool. Thanks for your time

[D Fosse]

Unless they have tweaked this adjustment to take the color space into consideration (which I don't expect), the answer is no. If it's mathematically the same adjustment, then no, they will not be the same. There will be a particularly significant difference between ProPhoto (gamma 1.8) and the others (+/- gamma 2.2).

I'm a photographer, coming from analog/darkroom, and I've always been very wary of trying to translate one technology into another. I've always worked from the assumption that analog/optical is one thing, digital data another. The aim is to get visual equivalents.

Because of that, I have always disliked the Brightness/Contrast and Exposure adjustments. They seem poorly defined, if at all, more like approximations to something people are "used to". I use Levels and Curves, I know what I get with those.

That said, if you need this kind of precise control, there is no way around a high-end monitor. By that I mean Eizo Coloredge or NEC Spectraview. With the very high precision you get with the integrated calibration/profiling software, you can be confident that what you see on screen is a correct representation of the file.

[Guillaume5FC6]

I work in video games as a concept artist. I have to work in 3 differents mediums. Real paint, gamma space with photoshop and game engine/ 3D packages that compute light linearly. I'm trying to bridge those a bit. Most digital artists on my team are at lost when it comes to gamma, linear light etc...   

Thanks for your time,

[TheDigitalDog]

The "tool" called "Exposure" in Photoshop and elsewhere has nothing to do with exposure which only takes place at capture and is an attribute of only the aperture and shutter that affects the amount of light striking the sensor. Forget the term exposure here and what takes place on gamma encoded images in Photoshop, in an RGB Working Space. This has nothing to do with measuring light or the amount of light striking a sensor or film. Just as the blend mode called Luminosity in Photoshop has nothing to do with luminosity.

[Guillaume5FC6]

Thanks, I totally see what you mean and you are probably 100% right. I knew 8bit & 16bit were gamma based and their math was done according to the color space you were using. But my understanding was that for a few features Photoshop would convert the gamma RGB to linear RGB data before applying the adjustment and then reconverting to your working color space. I must I've missed something somewhere.

I sometimes have to work with free floating 3D render passes in .EXR.  I thought the 32bit/channel mode was making calculations in RGB linear. I used Photoshop because it was convenient. Should, I do it in a real compositing software?

Thanks for your time.

[TheDigitalDog]

Bit depth and gamma encoding are utterly separate attributes. 

[Lumigraphics]

I recommend reading the Adobe Help pages regarding Exposure adjustment. They have good information 🙂

https://helpx.adobe.com/photoshop/using/color-adjustments.html#adjustments_panel_overview

https://helpx.adobe.com/photoshop/key-concepts/exposure.html

[Stephen_A_Marsh]

My understanding has always been that the "exposure" adjustments slider in Photoshop is for a HDR image workflow, 32 bpc, which is a linear gamma 1.0 workspace. Therefore, it is not surprising that gamma encoded spaces that have a 1.8 or 2.2 or another gamma will respond differently.